USGS Home
Contact USGS
Search USGS

U.S. GEOLOGICAL SURVEY
Scientific Investigations Report 2008–5169

Back to Table of Contents

Results and Discussion

Samples were selected to capture as wide a variety of materials (based on visual inspection) as was practical. Hence, measured core-sample properties were highly variable (table 2). Figure 4 shows the ranges in K_sat values measured in this study compared with those used in the development of the Winfield (2005) property-transfer model and measured by Pudney (1994), and those from Ackerman and others (2006), where horizontal K_sat values for each of the three hydrolgeologic units were estimated. These estimates are from (1) average linear ground-water velocities from numerous studies of atmospheric tracers, long term monitoring of contaminant movement in the aquifer, and knowledge of hydraulic gradients and effective porosities and (2) aquifer tests.

The values determined by Ackerman and others (2006) are much higher generally than the range of values for sediments because they represent the weighted average of sediment and basalt, and reflect a much larger scale of measurement. The ground-water flow model calibration process tends to assign lower values in areas where sediment proportions are greater than 11 percent in the upper part of the aquifer (Welhan and others, 2006). Welhan and others (2006) indicated that a more sophisticated scaling process based on sediment abundance would lead to improved estimates of horizontal K_sat values for the hydrogeologic units. This study provides data for deep aquifer sediments that can be used in scaling the composite K_sat values.

Laboratory-measured K_sat valuesvary over six orders of magnitude (fig. 4) with an average value of 8.78 × 10^-5 cm/s and a standard deviation of 1.45 × 10^-4. The average ρ_bulk was 1.64 g/cm³ with a standard deviation of 0.28. One sample from hole M2051 at 214.5 m depth had an unusually low ρ_bulk value (1.02 g/cm³) and also the highest measured K_sat value (4.78 × 10^-4). Excluding this anomalous sample, the average ρ_bulk was 1.71 g/cm³. The sample from hole M2050a at 362.7 m depth, which had a low K_sat value (1.25 × 10^-9 cm/s), was too consolidated to be disaggregated for particle-size analysis.

Particle-size distributions for nine of the core samples are shown in figure 5. Particle-size classes are listed in table 3 for the additional 67 bulk samples.

K_sat values for the 78 unsaturated zone sedimentary interbed samples used by Winfield (2005) in model development ranged from 1.1×10^-8 to 8.2×10^-2 cm/s with an average ρ_bulk of 1.46 g/cm³ (fig. 4). K_sat values for the seven deep samples used by Pudney (1994) ranged from 3.0×10^-7 to 1.4×10^-5 cm/s with an average ρ_bulk of 1.85 g/cm³ (fig. 4). Generally, K_sat values, as well as particle size statistics, show no consistent trend with depth; however, deeper interbeds tend to have higher ρ_bulk values. Figure 6 shows the difference in ρ_bulk values for deep samples (greater than 200 m depth from this study and Pudney 1994), and shallow samples (less than 60 m depth from Perkins and Nimmo, 2000; Perkins, 2003; and Winfield, 2003). The Man-Whitney rank-sum test (Zar, 1996) was used to determine that the difference in ρ_bulk between the shallow and deep samples statistically is significant. The calculated test statistic (Z) was 5.488, compared with the critical two-tailed value of 1.645 with an α level of 0.05. The samples from greater than 200 m depth indicate no strong linear trends between K_sat and bulk properties (ρ_bulk and median particle diameter); however, regression trends are slightly stronger for the Pudney (1994) data set than the data set from this study (fig. 7).

As shown in figure 8, with the exception of the lowest measured value, the Winfield model under predicts K_sat (AE value of 0.1110). The Pudney model (based on a combination of surficial and deep samples) over and under predicts about equally (fig. 8) with slightly more under predictions (AE value of 0.0009). Also shown in figure 8 are the predictions based on a linear fit to the K_sat and ρ_bulk data for the 10 core samples used in this study (labeled linear fit 1) and a linear fit to the data from this study combined with the 7 deep samples (labeled linear fit 2) from Pudney (1994). The small measurement error associated with the K_sat values (about 10 percent) has no influence on the relations examined here; error bars are imperceptible on figure 8. As listed in table 4, a linear fit to all available data from greater than 200 m gives the lowest RMSE value result, slightly lower than the Winfield and Pudney models. Predictions from the linear fit to the K_sat and ρ_bulk data for the 10 core samples from this study gives the highest RMSE value. The systematic under prediction by the Winfield model indicates that an adjustment could be made for samples with higher ρ_bulk or other systematic differences. Because the aquifer samples are affected by different processes by virtue of being deep as well as saturated, the addition of explanatory variables that include mechanical, chemical, and mineralogical parameters might yield better predictions.

Back to Table of Contents